Sorry; I see that the reply function on the forum didn't include my previous thread of emails.  I'll add it below for more context.  

So basically, I've done all of the steps you mentioned up through my group-level analyses.  I now have my tstats output along with the various other types of output from randomize (tfce p files, etc.) and am ready to run fdr on my uncorrected-p file.  

My question is whether I should be using masks of the activation and deactivation portions of the image to threshold the output file two different times, getting the activation and deactivation results separately (one fdr run with -m activationmask, and one fdr run with -m deactivationmask), or if I should be running fdr on the output file just one time and then using the masks of activation and deactivation afterward to see whether the thresholded results reflect activation or deactivation.  I'm not sure what the -m option is intended for.

Thank you,

Leslie

On Mar 23, 2012, at 11:16 AM, Christine Zakrzewski wrote:

Hi. I am not familiar with the conversation you had with Christian, but can provide some advice with regards to your question.
The order of operations would be--create within-subject mask (voxels common to all images in epi timeseries), within-subject analyses (deconvolution/ica), create group mask (voxels common to all individual within-subject masks), group-level analyses, multiple-comparisons correction (fdr, clusterize).
I hope this helps you.
Chrissy
 

> Date: Thu, 22 Mar 2012 20:18:39 +0000
> From: [log in to unmask]
> Subject: Re: [FSL] Dual-regression & contrast masking
> To: [log in to unmask]
> 
> Thanks, Christian. That does help. 
> I have a follow-up/related question.
> 
> I am using the fdr command to threshold my group-difference contrasts. I see that there is a masking option in the fdr command. Should I be including my activation mask (or deactivation mask, as it were) within the fdr command itself, or should I just be multiplying the fdr-sans-mask output by my activation (or deactivation) mask afterward in order to understand the results (e.g., if fdr thresholding of my groupA>groupB contrast returns results that happen to be within the regions of my deactivation mask, this would indicate that groupB's deactivation in those areas was stronger than groupA's)? 
> 
> Leslie

Hi

You can do this manually, using fslmaths to multiply the randomise output by whatever mask you want. 

Hth
Christian

On 06.02.2012, at 07:59, Leslie Aderhold <[log in to unmask]> wrote:


Hello FSLers - 

I am trying to analyze a set of resting-state data for which I have two separate subject groups.  I've set up my design matrix and contrast files for the dual-regression script, but I'd also like to mask some of my contrasts in order to separate activation from deactivation in my group comparisons.  I can see how contrast masking is set up using FEAT, but can anyone tell me how to use contrast masking in the context of dual-regression?   


Any guidance would be wonderfully appreciated!  

Leslie Aderhold

Cognitive Psychology PhD student
Penn State University
[log in to unmask]